Biological Conservation 142 (2009) 1141–1153

Contents lists available at ScienceDirect

Biological Conservation

journal homepage: www.elsevier .com/ locate /biocon

The Brazilian Atlantic Forest: How much is left, and how is the remainingforest distributed? Implications for conservation

Milton Cezar Ribeiro a,*, Jean Paul Metzger a, Alexandre Camargo Martensen a, Flávio Jorge Ponzoni b,Márcia Makiko Hirota c

a Departamento de Ecologia, Instituto de Biociências, Universidade de São Paulo, Rua do Matão, 321, Travessa 14, 05508-900 São Paulo, SP, Brazilb Departamento de Sensoriamento Remoto, Instituto Nacional de Pesquisas Espaciais (INPE), Avenida dos Astronautas, 1758, 12227-010, São José dos Campos, SP, Brazilc Fundação SOS Mata Atlântica, Rua Manoel da Nóbrega, 456, 04001-001 São Paulo, SP, Brazil

a r t i c l e i n f o a b s t r a c t

Article history:Received 17 September 2008Received in revised form 10 February 2009Accepted 14 February 2009Available online 24 March 2009

Keywords:Atlantic ForestConservationLandscape ecologyFragment sizeConnectivityMatrix influenceEdge effectsCore area

0006-3207/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.biocon.2009.02.021

* Corresponding author. Tel.: +55 12 39227496; faxE-mail addresses: mcr@usp.br, miltinho_astronauta

The neotropical Atlantic Forest supports one of the highest degrees of species richness and rates of ende-mism on the planet, but has also undergone a huge forest loss. However, there exists no broad-scale infor-mation about the spatial distribution of its remnants that could guide conservation actions, especiallywhen systematic biodiversity data are not available. In this context, our objectives were to quantifyhow much of the forest still remains, and analyze its spatial distribution. We considered the entireBrazilian Atlantic Forest, and eight sub-regions, defined according to species distribution. The resultsrevealed a serious situation: more than 80% of the fragments are <50 ha, almost half the remaining forestis <100 m from its edges, the average distance between fragments is large (1440 m), and nature reservesprotect only 9% of the remaining forest and 1% of the original forest. On the other hand, our estimates ofexisting Atlantic Forest cover were higher than previous ones (7–8%), ranging from 11.4% to 16%. The dif-ferences among estimates are mainly related to our inclusion of intermediate secondary forests and smallfragments (<100 ha), which correspond to approximately 32–40% of what remains. We suggest someguidelines for conservation: (i) large mature forest fragments should be a conservation priority; (ii) smal-ler fragments can be managed in order to maintain functionally linked mosaics; (iii) the matrix surround-ing fragments should be managed so as to minimize edge effects and improve connectivity; and (iv)restoration actions should be taken, particularly in certain key areas. The clear differences in the amountremaining and its spatial distribution within each sub-region must be considered when planning for bio-diversity conservation.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Landscape structure parameters have been recognized as usefulbiodiversity surrogates, and are used in different steps of conserva-tion planning (Williams et al., 2002; Lindenmayer et al., 2008).Some rules of thumb are employed within a landscape perspective,such as the conservation of large fragments with high structuralconnectivity, whether provided by corridors, stepping stones, orhigh permeability of the surrounding matrix (Umetsu and Pardini,2007; Umetsu et al., 2008; Uezu et al., 2008; Fonseca et al., 2009;Pardini et al., 2009; Vieira et al., 2009); as well as targeting thepreservation of as much as possible of all natural landscape heter-ogeneity (Forman and Collinge, 1997; Haila, 2002; Fischer et al.,2006; Lindenmayer et al., 2006; Metzger, 2006). Although theyhave some limitations, such as not considering how differentspecies perceive features of the landscape (e.g., functional

ll rights reserved.

: +55 11 30918096.@yahoo.com.br (M.C. Ribeiro).

connectivity), landscape structure parameters can be particularlyuseful to establish general guidelines for conservation planningwhere broad-scale species inventories and biodiversity distribu-tion patterns are still unavailable (Fairbanks et al., 2001; see Ueha-ra-Prado et al., 2009 for a example of ecological indicators), whichis the case for most tropical areas.

The Atlantic Forest was one of the largest rainforests of theAmericas, originally covering around 150 million ha (Fig. 1), inhighly heterogeneous environmental conditions. Its latitudinalrange is around 29�, extending into tropical and subtropical re-gions. The wide longitudinal range is also important in producingdifferences in forest composition, because of the decreased rainfallaway from the coasts. Coastal areas receive large amounts of rainyear-round, reaching more than 4000 mm, while inland forests re-ceive around 1000 mm/year (Câmara, 2003). These geographicalcharacteristics, combined with the large altitudinal range, have fa-vored high diversity and endemism, including more than 20,000species of plants, 261 species of mammals, 688 species of birds,200 species of reptiles, 280 species of amphibians, and many more

Fig. 1. Biogeographical sub-regions (BSRs) based on the main areas of endemism of birds, butterflies and primates as proposed by Silva and Casteleti (2003). Abbreviations forBrazilian states names are: AL = Alagoas, BA = Bahia, CE = Ceará, ES = Espírito Santo, GO = Goiás, MA = Maranhão, MG = Minas Gerais, MS = Mato Grosso do Sul,PE = Pernambuco, PB = Paraiba, PI = Piauí, PR = Paraná, RN = Rio Grande do Norte, RS = Rio Grande do Sul, SC = Santa Catarina, SE = Sergipe, SP = São Paulo, TO = Tocantins.

1142 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

species that still require scientific description (Goerck, 1997; Mit-termeier et al., 1999; Silva and Casteleti, 2003). The Atlantic Forestflora and fauna may include 1–8% of the world’s total species (Silvaand Casteleti, 2003).

Most of the remaining Atlantic Forest exists in small fragments(<100 ha; Ranta et al., 1998) that are isolated from each other andare composed by second-growth forests in early to medium stagesof succession (Viana et al., 1997; Metzger, 2000; Metzger et al.,2009). The few large fragments survived in locations where thesteep terrain made human occupation particularly difficult (Silvaet al., 2007). This present-day fragmentation has led to a large pro-portion of the forest’s vast biodiversity being threatened to extinc-tion; for example more than 70% of the 199 endemic bird speciesare threatened or endangered (Parker et al., 1996; Stotz et al.,1996; Goerck, 1997).

In these heterogeneous and highly diverse forests, which stillhold many still-unknown species (Lewinsohn and Prado, 2005),biodiversity inventories are complex, expensive and time-consum-ing (Gardner et al., 2008; but see Uehara-Prado et al., 2009). De-spite the large amount of biological data generated in theAtlantic Forest region in recent decades (Silva et al., 2004; Silvaand Casteleti, 2003), the lack of standardized inventory protocolsand sampling efforts with poor spatial distribution have resultedin significant geographical data gaps, making it particularly diffi-cult to use this information for conservation planning by the usualmethods (see Margules and Pressey, 2000; Groves et al., 2002). Atlocal scales, enough biological data is available for some areas tosupport conservation plans, but great difficulties arise in planningconservation actions for large regions. Moreover, most of the dataare insufficient to properly support conservation planning, and

thus, abiotic surrogates such as landscape structure parametersare in most cases the only alternative (Metzger et al., 2008). In thiscontext, the amount of habitat and fragmentation, which are keyfactors for biodiversity conservation (Wilcox and Murphy, 1985;Fahrig, 2003), are important variables to be considered in land-scape planning and management for biodiversity conservation.

Despite the potential of using landscape structure parameters inconservation planning, information on landscape structure in theAtlantic Forest is only available for small regions (<300,000 ha,Jorge and Garcia, 1997; Viana et al., 1997; Ranta et al., 1998).The only data available for the entire Atlantic Forest region is thepercentage of the remaining forest, but even in this case thereare huge discrepancies among the different methods employed(e.g., 7–8% of the forest remains according to SOS Mata Atlântica/INPE, 1993, 2000 and Galindo-Leal and Câmara, 2003; 10.6%according to SOS Mata Atlântica/INPE, 2008; and 27% accordingto IESB et al., 2007; Cruz and Vicens, in press).

We analyzed for the first time the spatial distribution of all theremaining Brazilian Atlantic Forest, in order to provide preciseinformation about how much forest is left and how this forest isspatially arranged. We calculated parameters such as fragment size,amount of edge area, isolation, structural connectivity, and distanceto conservation reserves of all existing fragments of the BrazilianAtlantic Forest region (ca. 245,000 fragments). These spatial analy-ses were performed on a multi-scale approach (Urban, 2005), in or-der to facilitate the biological interpretation of the landscapeindices (within a perspective of ecologically scaled landscape indices;following Vos et al., 2001). This approach ensures a variety of bio-logical behaviors, considering for example a wide range of sensitiv-ity to gap-crossing abilities and to edge distance influences.

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1143

Additionally, to refine our understanding about the Atlantic Forestspatial structure, considering its different regional contexts, weperformed the same analyses dividing the region by biogeographi-cal sub-regions (BSRs). The implications of the observed spatial pat-terns are discussed with regard to future conservation andrestoration priorities for the entire Atlantic Forest in Brazil.

2. Methods

2.1. Study region and biogeographical sub-regions (BSRs)

The Atlantic Forest originally extended from 3�S to 31�S, andfrom 35�W to 60�W, covering 148,194,638 ha (Fig. 1), mainlyextending along the Brazilian coast (92%), but also reaching intoParaguay (Cartes and Yanosky, 2003; Huang et al., 2007) andArgentina (Giraudo, 2003). The forest encompasses 17 Brazilianstates, and is narrow in the north and wider in the south. It hascomplex boundaries with other types of formations such as thepampas in the south and the drier inland formations, such as theBolivian Chaco and the Pantanal (west–southwest), the Cerrado(South American savanna, west) and the Caatinga (northwest).

We analyzed 139,584,893 ha (�94%) of the original BrazilianAtlantic Forest region, based on the extent defined in Brazilian leg-islation (Federal Decree No. 750/93 and Atlantic Forest law No. 11428, of December 22, 2006), and slightly expanded according to thedelimitation of BSRs by Silva and Casteleti (2003). The geographi-cal, historical and relief complexities observed in the Atlantic For-est region generate a scenario in which species are nothomogeneously distributed, but rather are grouped in differentBSRs (Silva et al., 2004; Silva and Casteleti, 2003). In order to prop-erly approach this complexity, we chose to conduct our analyses attwo different geographical scales: (i) the entire Atlantic Forest; and(ii) by the BSRs proposed by Silva and Casteleti (2003). Theseauthors defined five centers of endemism (Bahia, Brejos Nordesti-nos, Pernambuco, Diamantina and Serra do Mar) and three transi-tional regions (São Francisco, Araucaria and Interior Forests), basedon bird, butterfly and primate distributions (Fig. 1). For generaldescriptions we used all eight BSRs. For structural landscape anal-yses, the Brejos Nordestinos sub-region (1,251,783 ha) was ex-cluded because of its naturally scattered distribution andrelatively small area (<1% of the area studied). A detailed spatialanalysis is thus provided for seven different BSRs (Table 1). Withinthese regions, the Interior is the largest (49%), followed by theAraucaria (17%), and then by the Bahia, Serra do Mar and São Fran-cisco regions with about 8–9% each.

2.2. Forest cover

Forest cover analyses were based on an Atlantic Forestvegetation map (reference year 2005; www.sosma.org.br and

Table 1Extent of the Atlantic Forest in the biogeographical sub-regions (BSRs, as proposed by Silv

BSR Abbreviation Atlantic Forest domain

Area (ha)

Araucaria arauc 25,379,316Bahia bahia 12,241,168Brejos Nordestinos brejo 1,251,783Diamantina diama 8,289,516Interior inter 72,784,790Pernambuco perna 3,893,730Serra do Mar semar 11,413,471São Francisco sfran 12,940,866

Total 148,194,638

a Percentage of mapped area for each BSR in relation to total mapped area.

www.inpe.br) produced by the SOS Mata Atlântica/INPE (2008).The map was projected to the Albers projection and the SouthAmerica 1969 datum to assure accurate area calculation for largeregions such as this one. These institutions have been mappingthe Atlantic Forest cover since 1986, at 5-year intervals (SOS MataAtlântica/INPE, 2000 and 2008). The map used for the analysis wasconstructed by visual interpretation of TM/Landsat-5 (TM) andCCD/CBERS-2 (CCD) images from 2005 (with a few images from2004 when cloud cover did not allow image acquisition in 2005),viewed as color compositions on a digital orbital image mosaicwith bands TM3 (red region in the blue filter), TM4 (near infrared,NIR, in the red filter) and TM5 (short wavelength infrared, SWIR, inthe green filter) for Landsat-5, CCD-2 (green in blue filter), CCD-3(red in green filter) and CCD-4 (NIR in red filter) for CBERS-2. Thismap shows three main vegetation classes, grouping several physi-ognomically and floristically distinct forests: mangroves, ‘‘rest-inga” (lowland forests on sandy soils near the coast) and forests(including coastal forests, Araucaria mixed forests, and semi-decid-uous forests; Oliveira-Filho and Fontes, 2000). The three classeswere included in the forest cover analysis, but mangroves and res-tingas were not considered in the configuration analyses becauseof their small spatial extent (�4% of the remaining area) and bio-logical differences. The mapping scale was 1:50,000 in vector for-mat, which was then converted to raster, with a 50 m spatialresolution (60,000 � 48,000 cells), in order to improve metricscomputation.

The ‘‘forest” class included secondary forests in intermediate toadvanced successional stages. The distinction between old growthand secondary forest is particularly difficult for the entire AtlanticForest region because information about forest age is very scarceand available only at local scales. The old history of disturbancesin the region ended in a lack of good age estimation, especially forold forests which regenerated before the 1970’, when satelliteimages started to be available. We are aware that forest definitioncould originate differences in forest cover and configuration results,and thus we opted to consider as forest areas which have an arbo-real structure as seen by TM and CCD orbital images. This definitioncorresponds to forests >15 years of regeneration, with dense arbo-real vegetation and canopy height >10 m (Teixeira et al., 2009).

2.3. Map quality

At the end of the visual classification procedure, the preliminarythematic map was reviewed by vegetation experts from each Bra-zilian state located in the Atlantic Forest region. Fieldwork was alsoconducted in order to resolve questions of interpretation.

To assess the accuracy of the final map, we superposed the SOSMata Atlântica map on 10 reference cover maps (Table 1S), whichwere produced with high spatial accuracy (scales ranging from1:10,000 to 1:50,000) and extensive field checking. Of these maps,

a and Casteleti, 2003), and area mapped by SOS Mata Atlântica/INPE (2008).

Mapped by SOS Mata Atlântica/INPE (2008)

Area (ha) % mapped % of total mapped a

25,379,316 100 1712,241,168 100 8

85,249 7 18,200,259 99 6

68,417,731 94 493,132,167 80 3

11,413,471 100 810,715,533 83 9

139,584,893 94 100

1144 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

eight were available for the state of São Paulo (covering areas from10,400 to 24,800,000 ha), one for the state of Minas Gerais(143,900 ha), and another for the state of Paraná (748,500 ha).The 10 selected regions have very different reliefs and climates,but were limited to the forest pattern observed in southeast Brazil.As a consequence, our accuracy analysis may be biased towards thepatterns that occur in this area. We estimated the Kappa (Landisand Kock, 1977) and the G (percentage of pixels correctly classi-fied) statistics to check map accuracy (Table 1S). The Kappa valuesranged from 0.167 to 0.818 (mean = 0.486). The SOS Mata Atlânticamap showed low accuracy for only the region of one referencemap. For the other nine reference maps the analyzed map was clas-sified as acceptable (n = 3), intermediate (n = 4), high (n = 1) andvery high (n = 1) accuracy (according to the categories defined byLandis and Kock, 1977). To arrive at an overall statistic, we calcu-lated a Kappa value weighted for the area of the reference maps,and obtained K = 0.4 (i.e., acceptable according Landis and Kock,1977 classification).

We also estimated the errors of commission and omission, tocapture the bias of our map. The commission error ranged from0.2% to 8% (mean = 3%), with an area weighted mean of 2.9%. Theomission error ranged from 3% to 89% (mean = 49%), with an areaweighted mean of 37%. These results indicate that the map usedfor the analysis tends to underestimate the actual remaining vege-tation. This underestimation may occur because (1) early succes-sional stages are poorly mapped and (2) it is difficult to correctlymap the small fragments (<30 ha). We also noticed that for someregions, the map does not accurately show remnants on slopes,which resulted in some confusion between ‘‘forest” and Eucalyptusssp. plantations or early successional vegetation stages. Riparianforests were also poorly mapped, either because of their narrowshape or their overall small area. Despite these errors, the overallaccuracy (G) ranged between 76% and 97%, which is an acceptablerate for maps at this large regional scale.

2.4. Forest configuration indices

Forest cover and configuration metrics were computed for theentire study region and for each of the different BSRs using GRASS6.3 (Neteler and Mitasova, 2008; http://www.grass-gis.org), withsome procedures done on ArcGis 9.2 (Esri, 2007) and Erdas 9.1(Leica, 2006). The R language version 2.7.1 (R Development CoreTeam, 2008) was used for all data processing. We selected five con-figuration metrics that could be easily employed in conservationplanning: fragment size, edge area, connectivity, isolation and dis-tance to nature reserves (Table 2). For edge area (and core area),connectivity and isolation, we performed a multi-scale approach(Urban, 2005) in order to consider different species’ perception oflandscape structure.

Table 2Landscape metrics used to analyze the Atlantic Forest configuration.

Index Explanation

Fragment sizedistribution

Number of fragments and percentage of forest cover for different si

Edge area Percentage of area submitted to edge effects for different edge widt

Connectivity Area of functionally connected fragments considering different distfragment linkage

Mean isolation Mean isolation of random points to the nearest forest fragment. Toeffect of small fragments in estimating isolation, the smallest fragmsuccessively removed.

Distance fromNatureReserves

Distance of any given forest pixel to the nearest nature reserve

Fragment size distributions allowed us to account the forestamount and number of fragments for different classes of size (Ta-ble 2). Edge area was computed as the amount and percentage offorest area submitted to edge effects for different edge widths.All forest pixels with distances higher than an edge width level (Ta-ble 2) were classified as core area.

Connectivity metrics were computed based on the graph the-ory (Urban and Keitt, 2001; O’Brien et al., 2006; Minor and Ur-ban, 2007; Fall et al., 2007), which is a method of measuringthe functionally connected clump of fragments based on somesimple linkage rules (Urban and Keitt, 2001). As linkage rulesfor our analysis we considered the distance among fragmentsreflecting different gap-crossing capacities (Martensen et al.,2008; Boscolo et al., 2008; Awade and Metzger, 2008; Table 2).The connectivity index was then calculated as the sum of theareas of clumped fragments, which can be interpreted as thefunctional available area (Martensen et al., 2008; Metzger et al.,2009). After generating connectivity maps, we computed the ex-pected cluster size as the mean clump size for each functionaldistance. The highest cluster size was also identified for the en-tire study region, as well as for each sub-region and for eachfunctional distance.

To estimate mean isolation, we proposed an index adaptedfrom the ‘‘Empty Space Function”, which is a spatial point patternanalysis (Baddeley and Turner, 2005; Fortin and Dale, 2005), thatrepresents the mean isolation of a given random pixel from anyforested one. We randomized 1,000,000 points over the entireAtlantic Forest region, and a distance map was generated for allfragments. We then successively removed the smaller fragmentsin several steps (Table 2) and computed the distance to the near-est forest in each step. These values represent the isolation of for-est areas, but are particularly useful in providing insights aboutthe importance of the smaller fragments (or the capacity of thespecies to use these small fragments as stepping stones; Uezuet al., 2008).

To assess the amount of Atlantic Forest protected by the Brazil-ian Protected Area network, we superposed on the remnants map,a map of Nature Reserves (strictly protected areas classified as ‘‘pro-teção integral”; MMA, 2007), which comprises 249 reserves(2,260,350 ha). The distance from nature reserves was also calcu-lated for each forest pixel (see Table 2 for classes of distance fromreserves).

3. Results

3.1. Forest cover

Of the total mapped area (139,584,893 ha), 15,719,337 ha offorest (11.26%) and 658,135 ha (0.47%) of restinga and mangrove

Classes or rules

ze classes Fragment size classes (ha): <50, 50–100, 100–250, 250–500, 500–1000, 1000–2500, 2500–5000, 5000–10,000, 10,000–25,000,25,000–50,000, 50,000–100,000, 100,000–250,000, 250,000–500,000 and 500,000–1200,000

hs Edge widths (m): <50, 50–100, 100–250, 250–500, 500–1000,1000–2500, 2500–5000 and 5000–12,000

ance rules for Linkage distances (m): 0, 100, 200, 300, 400, 500, 1000 and 1500

analyze theents were

Size of the small fragments removed (ha): 0 (i.e., no fragmentremoved), <50, <100, <150, <200, <350 and <500

Distance classes (m): 0 (i.e., inside a Nature Reserve), <200, 200–600, 600–1000, 1000–2000, 2000–5000, 5000–7500, 7500–10,000,10,000–25,000, 25,000–50,000 and >50,000

Table 3Remaining Atlantic Forest in each biogeographical sub-region (BSR), with its area in ha and percentage. Data were obtained by superposing the map generated by SOS MataAtlântica/INPE (2008) and the boundaries of BSRs adapted from Silva and Casteleti (2003).

BSR Remaining forest Remaining restinga/mangrove Total remaining Atlantic Forest

Area (ha) %a Area (ha) %a Area (ha) %a

Araucaria 3,202,134 12.6 3,202,134 12.6Bahia 2,047,228 16.7 115,059 0.9 2,162,287 17.7Brejos Nordestinos 13,656 16.0 13,656 16.0Diamantina 1,109,727 13.5 1,109,727 13.5Interior 4,807,737 7.0 32,451 4,840,188 7.1Pernambuco 360,455 11.5 19,363 0.6 379,818 12.1Serra do Mar 3,678,534 32.2 491,263 4.3 4,169,797 36.5São Francisco 499,866 4.7 499,866 4.7

Total 15,719,337 11.26 658,135 0.47 16,377,472 11.73

a Percentages are in relation to the BSR area.

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1145

vegetation (Table 3) still remain. Thus, 88.27% of the original Atlan-tic Forest has been lost, and only 11.73% of the original vegetation(16,377,472 ha) remains (Fig. 2 and Fig. 1S[a–g]). Considering theestimated commission and omission errors for the map analyzed(Table 1S; see Map quality section for details), we consider thatthe actual remaining vegetation might range from 11.4% to 16.0%in the entire Atlantic Forest region.

The best-preserved BSR is the Serra do Mar, which holds 36.5%of its original vegetation, followed by the Bahia (17.7%) and BrejosNordestinos (16%) regions. In contrast, the São Francisco region hasonly 4.7% of forest cover, and the Interior Forest, 7.1% (Table 3).However, in absolute terms, more than half of the remaining forestis located in the Serra do Mar and Interior Forest regions, whereasless than 15% is located in the Diamantina, São Francisco, BrejosNordestinos and Pernambuco regions (Table 3).

Fig. 2. Remaining forest in the Brazilian Atlantic Forest region (source: SOS Mata A

3.2. Number of fragments and size distribution

The Atlantic Forest is currently distributed in 245,173 forestfragments. The largest fragment is located in the Serra do Mar,mainly along the coastal mountains of the state of São Paulo, andextends from the state’s southern border northwards into thesouthern part of the state of Rio de Janeiro. This single fragmentcontains 1,109,546 ha of continuous forests, which represents 7%of what remains (Fig. 2 and Fig. 1S-f). The second- and third-largestfragments are also located in the Serra do Mar, and contain508,571 ha (coastal zone of Paraná state) and 382,422 ha (coastalzone of Santa Catarina state). Altogether, the three largest frag-ments account for more than 2 million ha, i.e., more than 13% ofthe remaining forest. In contrast, 83.4% of the Atlantic Forest frag-ments (204,469 fragments) are smaller than 50 ha, and together

tlântica/INPE, 2008). See Fig. 1 for the abbreviations of Brazilian states names.

Fig. 3. Distribution of remaining forest fragment sizes in the full extent of theAtlantic Forest region. %A: percentage of total area; %NP: percentage of number offragments.

0

5

10

15

20

25

30

250- 5001000

1000-2500

2500-5000

5000-12000

Edge Distance (m)

%

Edge Distance (m)

Cum

ulat

ive

area

(%)

2550

7510

0

50 100 250 500 1000 2500 5000 12000

0-50 50-100 100- 250 500-

A

B

Fig. 4. Cumulative (A) and per class (B) area under edge effect at different depthsfor the remaining Brazilian Atlantic Forest. Edge depths of 100 and 250 m arehighlighted in (A).

1146 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

they account for 20.2% of the total forest remnants (ca.3,178,030 ha; Fig. 3) according to our estimates. Fragments smallerthan 250 ha represented more than 97% of the total number, andaccounted for almost 42% of the total forest area. In contrast, only0.03% (77 fragments) are larger than 10,000 ha, and together theseinclude almost 4 million ha.

Small fragments (<50 ha) are, by far, the largest part of thenumber of remnants in all BSRs. The distribution of fragmentsaccording to their size followed an inverted ‘‘J” shape (Fig. 2S).However, the Serra do Mar region, in addition to having manysmall fragments (�79% of the fragments is <50 ha), has much ofthe forest existing as large fragments (>50,000 ha), which repre-sent more than 50% of the forest cover in the region. This regionis the only one with a fragment larger than 1 million ha in size,which is located along the coastal mountains of São Paulo. Theother regions do not contain any fragment larger than250,000 ha, and only the Araucaria Forest (n = 4) and the InteriorForest (n = 1) have forest fragments larger than 50,000 ha(Fig. 2S): the inland forests of Santa Catarina, including the São Joa-quim National Park, and the Iguaçú National Park, respectively. Inthe Bahia region, the largest fragment covers approximately29,000 ha, while in the São Francisco and the Pernambuco regions,none exceeds 10,000 ha (see Fig. 2S[a–g] for detail); and in theDiamantina, none is larger than 25,000 ha.

3.3. Core and edge area

Of the total forest area remaining, 73% is located less than250 m from any non-forest area, and 46% is less than 100 m distantfrom edge (Fig. 4). Only 7.7% is located farther than 1000 m fromany edge, and 12 km is the maximum distance from any non-for-ested area in the Atlantic Forest region.

A similar pattern was observed for the BSRs, where most of theforest area is less than 250 m from non-forest areas (Fig. 3S). In theInterior and Pernambuco regions, approximately 60% of the forestis less than 100 m from any edge, while São Francisco, Araucaria,Bahia and Diamantina have between 40% and 50% of their forestswithin 100 m from any edge. Only the Serra do Mar region showeda different pattern, where only 25% of the remaining forest is lo-cated less than 100 m from any edge. This is reflected in the high-est percentage of core-area forest in this BSR, with 256,040 ha atleast 2.5 km from the edges, and 56,993 ha at least 5 km fromthe edges. In addition to the large fragments of the Serra do Mar

region, the Iguaçu National Park is the only one that also has areasof forest that are 12 km distant from any edge.

3.4. Connectivity

For species that are not able to cross open areas, the averagefunctionally connected area is 64 ha (Fig. 5), while for those thatare able to cross 300 m it is 131 ha. The largest functionally con-nected cluster of fragments for species that are able to cross100 m, is comprised of the Serra do Mar and the nearby functionallyconnected fragments, which encompass more than 2,803,000 ha(18% of the remaining forest, Fig. 6) and stretches from the stateof Rio de Janeiro all the way south to the state of Rio Grande doSul. In the Bahia region, species that are able to make short cross-ings between fragments, such as 100 m, can reach a forest area ofmore than 50,000 ha (17% of the remaining forest in the region);whereas in Diamantina the gap that needs to be crossed to reacha functionally connected area of this size is 200 m (Fig. 4S-c). Longerdistances separate fragments in the other regions, such as 400 m toreach 50,000 ha in the São Francisco (Fig. 4S-g) BSR, and more than500 m in the Pernambuco to reach the same area (Fig. 4S-e).

3.5. Mean isolation

The mean isolation for the entire Atlantic Forest region was1441 m, with values ranging from a few meters to dozens of kilome-ters. The small fragments were particularly important in reducingisolation (Fig. 7). When we exclude the fragments <50 ha, the meanisolation increases to 3532 m. If fragments smaller than 200 hawere lost, the mean isolation would reach more than 8000 m.

Fig. 5. Expected cluster size (mean functional size; ha) for functionally connectedforest fragments estimated across varying functional linkage distances (m), for theBrazilian Atlantic Forest region.

Fig. 6. Highest functionally connected forest cluster (% of total remaining forest)estimated across varying functional distances (m), for the Brazilian Atlantic Forestregion.

Fig. 7. Influence of the smallest fragment size (ha) on the mean isolation (m)between fragments, for the Brazilian Atlantic Forest region, and for its biogeo-graphical sub-regions (BSRs). See Table 1 for BSRs abbreviations. Smallestfragments sizes: 0 ha (all fragments), 50 ha, 100 ha, 150 ha, 200 ha, 350 ha and500 ha.

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1147

Small fragments were important in reducing isolation in all re-gions. However, a gradient of importance could be seen, with theInterior and São Francisco BSRs being particularly affected by theexclusion of these small fragments, whereas in the Serra do Mara relatively low isolation is maintained, since the remaining forestexists as larger pieces (Fig. 7). The isolation in the São FranciscoBSR is the most-affected by the exclusion of small fragments, sinceisolation increases from 3.6 to 14.5 km when excluding fragmentssmaller than 200 ha. The second most-affected BSR is the Interior,where the mean isolation increases from 1344 m (without frag-ment removal) to 9112 m if we remove fragments <200 ha. Forother regions, exclusion of fragments of this size results in an iso-lation that ranges from 4182 m (in the Serra do Mar) to 7048 m (inthe Araucaria BSR).

3.6. Nature reserve cover and proximity

The total protected area within the Atlantic Forest region isapproximately 2.26 million ha, or 1.62% of the region (Table 4,Fig. 8). Nature reserves represent 14.4% of the remaining forest

cover, but they protect only 9.3% (Table 4) of this remaining forest,since other types of vegetation or land cover also occur withinthese reserves. All regions have a small percentage (Fig. 5S[a-g])of their areas covered by nature reserves. However, the Serra doMar has 25.2% of its remaining forest under protection, followedby the Interior (6.8%) and the Bahia (4.2%). All other regions haveless than 4% of their small amount of remaining forest under pro-tection (Table 4; Fig. 5S[a-g]). Given these facts, the Serra do Maraccounts for 63% of the total remaining forest under protection, fol-lowed by the Interior BSR (22%). Moreover, only 1.05% of the origi-nal forest cover is protected, and in most regions (except Serra doMar and Bahia) this percentage is <0.5%.

Some reserves are contiguous, and thus we could identify sevenlarge protected regions with areas of about 100,000 ha. Five are inthe Serra do Mar region: (1) Serra do Mar State Park and BocainaNational Park; (2) Jacupiranga State Park and Superagui NationalPark; (3) Paranapiacaba [Petar State Park, Intervales State Park,Xituê Ecological Station and Carlos Botelho State Park]; (4) Serrado Tabuleiro State Park and (5) Jureia [Banhados de Iguape Ecolog-ical Station, Jureia-Itatins Ecological Station, Itinguçu State Parkand Prelado State Park]. The other two regions are in the Interior(Iguaçú National Park) and the Diamantina (Chapada da Diamanti-na State Park) regions. Together they have a total area of1,212,800 ha, which encompasses 53.6% of the protected areas.Seventeen reserves range in size from 20,000 ha to 60,000 ha (intotal 585,120 ha; 26% of the protected areas); six of them are inthe Interior, five in the Serra do Mar, three in the Bahia, two inthe Araucaria, and one in the São Francisco BSRs.

Only 22.6% of the remaining forest is located within 10 km ofnature reserves, whereas 61% is farther than 25 km (Fig. 9). MostBSRs have a small amount of forests close (<10 km) to nature re-serves, whereas a large amount is more distant (>50 km, Table 4;Fig. 5S[a–g]). A different pattern is encountered in the Serra doMar, where 59% (2,163,163 ha) of the remaining forest is less than10 km, and 41% (1,515,371 ha) is farther away.

4. Discussion

Our results showed that: (i) there is more forest left thanpreviously estimated; (ii) most fragments are very small, less

Table 4Protected area and forest under protection for the Atlantic Forest domain, and within seven biogeographical sub-regions (all BSRs except ‘‘Brejos Nordestinos”) adapted from Silvaand Casteleti (2003).

BSR Area ofsub-region(a)(ha)

Protected area Remainingforest (c)ha

Protectedremainingforest (d)ha

Protected forest in relationto remaining forest withinsub-regions (d)/(c)%

Protected forest betweensub-regions(di)/S(d)%

Protected forest in relationto original forest withinsub-regions (d)/(a)%

(b)(ha)

(b)/(a)%

Araucaria 25,379,316 164,651 0.65 3,202,134 98,121 3.1 6.7 0.39Bahia 12,241,168 113,447 0.93 2,047,228 86,053 4.2 5.9 0.70Diamantina 8,200,259 151,412 1.85 1,109,727 12,451 1.1 0.9 0.15Interior 68,417,731 561,381 0.82 4,807,737 325,261 6.8 22.2 0.48Pernambuco 3,132,167 4314 0.14 360,455 3731 1.0 0.3 0.12Serra do Mar 11,413,471 1,201,848 10.53 3,678,534 926,184 25.2 63.3 8.11São Francisco 10,715,533 63,297 0.59 499,866 11,823 2.4 0.8 0.11

Total 139,499,644 2,260,350 1.62 15,705,681 1,463,622 9.3 100.0 1.05

Fig. 8. Nature reserves in the Brazilian Atlantic Forest region (source: MMA, 2007).

10000

100000

1000000

10000000

NR<20

0

200-

600

600-

1000

1000

-200

0

2000

-500

0

5000

-100

00

1000

0-25

000

2500

0-50

000

>5000

0

Distance from Nature Reserves (m)

Are

a (h

a in

log

scal

e) 9.3%

0.5%0.8% 0.7%

1.5%

4.0%5.9%

16.5% 19.3%

41.6%

Fig. 9. Remaining forest (area and percentage) within nature reserves (NR) andamount of forest per class of distance from nature reserves (m) for the BrazilianAtlantic Forest region.

1148 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

than 50 ha (ca. 83% of the total number of fragments); (iii)much of the remaining forest is close to forest edges (ca. 45%<100 m of the edges), indicating that matrix influences mayhave strong effects on many forest ecological processes (Umetsuand Pardini, 2007; Umetsu et al., 2008; Uezu et al., 2008;Fonseca et al., 2009; Pardini et al., 2009; Vieira et al., 2009);(iv) short gap-crossings (<100 m) through the matrix can behighly effective to increase the functionally connected area forforest species (Boscolo et al., 2008; Martensen et al., 2008);(v) small fragments (<200 ha) play a crucial role in reducingfragment isolation among larger fragments, suggesting that theyare highly important as stepping stones; and (vi) nature re-serves protect a small amount of the remaining forest in allthe BSRs (except in the Serra do Mar), and most of the remain-ing forest is distant (>25 km) from the existing nature reserves(61%; 9,564,900 ha).

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1149

4.1. How much forest is left?

Our study exposed the extreme degradation of the Atlantic For-est, where only 11.7% of the original vegetation remains(15,719,337 ha). This proportion might range from 11.4% to 16.0%if we consider errors of commission and omission. This estimatedarea is larger than the usual total given for the Atlantic Forest(7–8%; SOS Mata Atlântica/INPE, 1993, and 2000; Galindo-Lealand Câmara, 2003), but below the recent estimate of 27% by IESBet al. (2007) and Cruz and Vicens (in press). Differences amongthese estimates could be caused by several factors, including map-ping errors. However, our field survey demonstrated that the mapquality is acceptable for regional analyses. Apparently, one of themain factors causing these discrepancies in estimating the AtlanticForest cover is related to the criteria used to include secondary for-ests and small fragments.

IESB et al. (2007), Cruz and Vicens (in press) included very earlystages of succession, even without a forest structure present (veg-etation <2 m high; Cruz, personal communication). The initial eval-uations by the SOS Mata Atlântica Foundation (1986–2000) did notconsider regrowth; i.e., once mapped as deforested in a recent past(�50 years), an area was never again considered as forest (SOSMata Atlântica/INPE, 2008). They also did not consider forest frag-ments smaller than 100 ha. On their new report, when consideringfragments >3 ha and forests at an intermediate stage of regenera-tion, they obtained an estimation of the remaining forest coversimilar to ours (10.6%; SOS Mata Atlântica/INPE, 2008). Thus, thedifference between previous estimates by SOS Mata Atlântica/INPE(1993,2000) and the present estimate is basically composed ofintermediate secondary forests and/or small fragments, which cor-respond to approximately 32–40% of the total mapped vegetation(considering 8% and 7% of the forest cover previously mapped bySOS Mata Atlântica, respectively). The small difference betweenour estimates and that reported by SOS Mata Atlântica/INPE(2008) may be related to differences on the analyzed extent(slightly larger in our case to include the delimitations of BSRs bySilva and Casteleti, 2003) and because we did not excluded anyfragment from the analyses.

The Atlantic forest presents a highly dynamic forest cover dy-namic (Teixeira et al., 2009) and an old history of disturbance,where even the mature remnants were selective logged in a distantpast, sometimes in pre-European time (Dean, 1996; Câmara, 2003).This critical scenario of the Atlantic Forest with large areas of sec-ondary forests and small disturbed fragments is the usual patternin most tropical regions, where 50% of them are secondary or dis-turbed (Wright, 2005). The implications for species conservationare huge. Even if secondary forests can sustain a significant amountof biodiversity (Viana and Tabanez, 1996; Develey and Martensen,2006), many species need more pristine forest and large fragmentsto survive (Aleixo, 1999; Harris and Pimm, 2004; Develey and Mar-tensen, 2006; Laurance, 2007; Gardner et al., 2007; Barlow et al.,2007a,b), and the replacement of mature forests by secondary onesmay lead to the extinction of many species (Metzger et al., 2009).

In this landscape-structure scenario, several general guidelinescan be suggested to improve or stimulate forest-species conserva-tion in the Atlantic Forest region, particularly: (i) large mature forestfragments should be assigned a high conservation priority; (ii) smal-ler fragments could be used to form functionally linked mosaics; (iii)the matrix surrounding the fragments should be adequately man-aged to minimize edge effects; and (iv) restoration actions shouldbe put into practice, particularly in some key structural conditions.

4.2. Protecting large mature forest fragments

Particular attention should be paid to the protection of the lar-ger remnants, especially for their capacity to maintain larger pop-

ulations and for their better prospects of sustaining species overthe long term (Brooks et al., 1999; Lindborg and Eriksson, 2004).Only large fragments with mature forests are capable of preservingsensitive species, especially those with large area requirements(Ferraz et al., 2007) or with strict habitat requirements, whose sur-vival is particularly problematic in the present fragmented state(Aleixo, 1999). Old-growth forests (Laurance, 2007; Gardneret al., 2007) are especially important because, even in excellentregeneration conditions, this process requires several decades torestore a species composition comparable to a mature stage (Dunn,2004; Barlow et al., 2007a,b; Liebsch et al., 2008; Rodrigues et al.,2009). Moreover, the early stages of succession on abandonedlands are not protected by Brazilian environmental laws (Metzgeret al., 2009), and their suppression is common and part of the rapiddynamic of the landscapes.

In the present extremely fragmented and dynamic scenario ofthe Atlantic Forest, the importance of the last large forest remnantsincreases exponentially, and their management should be cau-tiously designed and their transformation into nature reservesshould be carefully considered. New conservation areas are mosturgently needed in Pernambuco, Diamantina and São Francisco,where nature reserves protect less than 3% of the remaining forest,even if the largest fragments are not so large (e.g., <10,000 ha). Pro-tection of at least 10% of the original habitat is recommended as aglobal strategy for conservation (Secretariat of the Convention onBiological Diversity, 2002), but even the Serra do Mar region doesnot at present fulfill this condition (8.1% of its original cover is pro-tected in nature reserves). This kind of conservation target is sub-jective, and the degree of protection should vary according to thesensitivity of the system (which may differ among regions), butit is clear that the proportion of forest cover under legal protectionis very low in all the BSRs, and must be enhanced.

Moreover, the largest fragments promote conservation of coreareas, which are particularly uncommon since more than 70% ofthe remaining forest is located less than 250 m from open land,and is thus subject to edge effects. Furthermore, large mature for-est fragments are vital for supporting seeds and allowing recoloni-zation of the small surrounding fragments, and can act as sourceareas for restoration programs (Rodrigues et al., 2009). The dis-tance (>25 km) of most of the remaining Atlantic Forest fragmentsfrom the existing nature reserves reduces their influence as stablesources of individuals and species for the surrounding smaller frag-ments. Forest regeneration in areas around nature reserves shouldbe stimulated because of their natural regrowth potential, increas-ing forest cover in these regions, and reducing the present unbal-anced distribution of forest cover in relation to the proximity tonature reserves.

Protecting the last largest blocks of forests of all BSRs shoulddefinitely be a conservation priority. However, only in the Arau-caria, Bahia, and especially the Serra do Mar regions can large frag-ments still be found. The Interior Forest also has a few largefragments, which together with the large tracts of forests stillexisting in Argentina (Giraudo, 2003) and Paraguay (Cartes andYanosky, 2003; Huang et al., 2007) constitute a better conservationprospect for this BSR. All the other BSRs lack large fragments,which may severely compromise species conservation (Silva andCasteleti, 2003). In these regions, alternative conservation prac-tices should be implemented for the existing remnants.

4.3. Creating functionally linked mosaics

The conservation of small fragments should not be neglected,because they constitute a large fraction of the remnants (83.4%with <50 ha), and are essential in enhancing connectivity betweenthe larger ones. The mean distance between fragments for theAtlantic Forest is around 1400 m, considering the small fragments.

1150 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

This is a considerable separation for most forest species (Lauranceand Gómez, 2005), which avoid edge areas (Hansbauer et al., 2008;Lopes et al., 2009) and in some cases do not even cross roads orsmall gaps in the canopy (Develey and Stouffer, 2001; Laurance,2004; Laurance and Gómez, 2005). As the distances between frag-ments increase, connectivity decreases, and individual crosses be-came less frequent (Hanski, 1994; Haddad, 1999). However,clusters of neighboring fragments (<200 m) that form large tractsof forest (>50,000 ha) are common (n = 12; 4,992,700 ha; 32% ofthe total remaining forest), and should be considered in conserva-tion policies as important potential mosaics for conservation. Theycan play an important role in animal movement through land-scapes, either functioning as stepping stones (Castellón and Siev-ing, 2005; Sekercioglu et al., 2006; Uezu et al., 2008; Boscoloet al., 2008), or forming networks of functionally connected areas,which could allow species to persist in disturbed landscapes (Mar-tensen et al., 2008). They can also act as stable sources of seeds andindividuals for nearby, smaller fragments.

Especially in cases where no large fragment is left, one option isto consider functionally linked mosaics of smaller fragments (Uezuet al., 2005; Martensen et al., 2008). This is especially the case inthe Pernambuco, Diamantina and São Francisco BSRs. These BSRsare naturally fragmented (Fig. 1), and the advanced forest destruc-tion intensified this pattern, increasing the number of small frag-ments, reducing the forest cover to <15% and thus threateningthe biodiversity that depends on this forest.

Matrix permeability is a key connectivity component (Fonsecaet al., 2009). Different regions of the Atlantic Forest have sufferedfor centuries under poor soil management and extensive destruc-tion and simplification of the forests, which has intensified in re-cent decades (Fig. 6S), with agricultural mechanization and theuse of pesticides and herbicides (Brannstrom, 2001; Duriganet al., 2007). Recent expansions of Eucalyptus (Bacha and Barros,2004; Baptista and Rudel, 2006; Fig. 6S) and sugar-cane planta-tions (Rudorff and Sugawara, 2007, Nassar et al., 2008; Fig. 6S)have caused huge social, economical and environmental impacts,and their influences in reducing patch connectivity have been sug-gested and need to be better understood (Fonseca et al., 2009).

Matrix management is especially important in a scenario whereshort movements through the matrix can promote fragment con-nections. Increasing matrix permeability can allow species to per-sist in fragmented situations and can be an interesting option forconservation of some species (Ricketts, 2001; Baum et al., 2004;Antongiovanni and Metzger, 2005; Umetsu et al., 2008; Uezuet al., 2008; Pardini et al., 2009; Vieira et al., 2009). Matrix man-agement could be employed in the Araucaria Forest, where func-tionally connected areas increase stability through improvedgap-crossing abilities; and in the Bahia region, where a large net-work of fragments are functionally linked by distances of 100 m ofopen areas. In all these cases, inter-forest management can im-prove connectivity. The implementation of forested pastures,where trees are scattered in different densities, or disposed in linesalong fences, has been suggested as one easy option to improveconnectivity in traditional pasture areas, while also promotingadditional economical gains to the landowner, as well as work inwind blocking (Harvey et al., 2004). The establishment of smallagroforest patches, with the presence of trees dispersed in agricul-tural fields is also suggested as important to improve landscapeconnectivity (Uezu et al., 2008). Moreover, sugar cane has beenexpanding over former pastures, where scattered trees were com-mon (Nassar et al., 2008). The suppression of these trees should beavoided, because they can have a disproportional effect in biodi-versity conservation (Harvey et al., 2004). Finally, different man-agement options in the plantations of exotic trees are alsopointed out as vital to enhance connectivity, as well as in harbor-ing some extremely demanding species, what could be severely

limited with traditional forest management practices (Fonsecaet al., 2009).

4.4. Reducing edge effects

Because the remaining Atlantic Forest is severely fragmented insmall patches edge effects increase in importance. Tropical species,particularly the strictly forest ones, are well known to be highlysensitive to edge alterations, especially because of their high nichespecialization (Kapos, 1989; Murcia, 1995; Hansbauer et al., 2008;Lopes et al., 2009). Kapos (1989), Laurance et al. (2007), Laurance(2008) showed that some edge effects in the Amazon Forest canextend as far as 300–400 m into the forest. In the Atlantic Forest,almost half of the remaining forest is located less than 100 m fromopen areas, and more than 70% of the remaining forest is locatedless than 250 m from open land, i.e., is subject to strong edge-effectinfluences.

Because most of the present forest is directly influenced bynearby land use, matrix influences should also be particularlyinvestigated and their management carefully conducted (Umetsuand Pardini, 2007; Umetsu et al., 2008; Uezu et al., 2008; Fonsecaet al., 2009; Pardini et al., 2009; Vieira et al., 2009). Forested oragroforestry matrixes are suggested as efficient in reducing micro-climate changes caused by edge conditions (Didham and Lawton,1999), and thus, to reduce edge-effect influences (Cullen Junioret al., 2004). In some cases, strip of trees are planted forming a buf-fer around the remaining patches in order to reduce edge effects(Cullen Junior and Fenimore, 2002; Cullen Junior et al., 2004).

4.5. Stimulating restoration actions in key structural conditions

The present small amount of remaining forest generates a sce-nario where rapid restoration actions should be undertaken to al-low species conservation in the near future (Rodrigues et al.,2009). In this case, careful site selection is necessary in order tomaximize restoration achievements. Priority should be given tothe bottlenecks of the large clusters of fragments. In the Serra doMar region, for example, small disruptions (<100 m) break aparta potential larger fragment of more than 2.8 million ha, and resto-ration of these connections should be a conservation priority. Inthis region, roads are also a key element in breaking apart largefragments, and mitigation policies should be analyzed and em-ployed to restore connectivity. Complementarily, efforts shouldbe allocated to create new reserves between the large fragmentsof the Serra do Mar, enhancing reserve sizes and width, especiallyin areas where forest areas are relatively narrow and disruption ofthese fragments could occur in the future.

Another key characteristic of the Serra do Mar region is thepresence of large fragments that extend into the Interior and Arau-caria regions. Maintenance or linkage restoration of these inter-re-gion fragments is also vital to maintain the biological evolutionaryprocesses, which could be particularly important in the presentscenario of climate changes. These linkages may be especiallyimportant in the case of the Araucaria BSR, since these inter-regionfragments are also the largest ones found in this BSR, and the re-gion has been undergoing profound impacts from the recentchanges in land use, with the expansion of Eucalyptus plantations(Bacha and Barros, 2004; Baptista and Rudel, 2006).

The northern Atlantic Forest is presently in worst state of con-servation compared to the southern part, except for the large mo-saic in the south of the Bahia region. This mosaic should be a keytarget for biological conservation, and the restoration of forest con-nections should be a main concern in this region. The absence offorest in the area between Bahia and the Serra do Mar makes it par-ticularly difficult for individual animals to pass between the south-ern and northern Atlantic Forests, which could severely impact

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1151

evolutionary processes and could be particularly deleterious in aclimate-change scenario.

Finally, restoration priority should also be given to link thesmaller fragments surrounding larger ones, especially in the caseswhere these large fragments are nature reserves. Enhancing theconnectivity between fragments, especially linking to a large onethat can act as a source of individuals, can improve conservationpossibilities in highly fragmented regions (Uezu et al., 2005; Mar-tensen et al., 2008).

5. Conclusion

This report quantifies for the first time the extremely degradedstate of the Atlantic Forest distribution, showing that most frag-ments cover less than <50 ha, almost half the remaining forest is<100 m from forest edges, and the present conservation networkis insufficient to support the long-term survival of this rich andendangered tropical forest. Urgent conservation and restorationactions should be implemented to mitigate this situation, basedon careful planning and with clear targets.

The management of the region as a whole must begin with thetransformation of the large mature forest tracts into conservationreserves and the reestablishment of key connectivity linkages,especially between the larger remnants. The matrix, mainly be-tween and surrounding these large fragments, is also importantto manage, since a large fraction of the forest is influenced by theclose proximity to edges. In sections where not a single large frag-ment remains, which is common in most sub-regions, manage-ment should focus on the reestablishment of functionallyconnected clusters of fragments, enhancing landscape connectiv-ity. In the present critical conservation scenario of the Atlantic For-est, every remnant is important for species conservation. The cleardifferences in the amount of forest remaining and how they are or-ganized in each sub-region must be considered when planning bio-diversity conservation.

Acknowledgments

Our profound thanks to the SOS Mata Atlântica Foundation andthe National Institute for Space Research (INPE) for allowing us touse their maps of remnants of the Atlantic Forest region. We alsothank J.M.C. Silva and C.H.M. Casteleti for providing the boundariesof the BSRs analyzed in this study. Our appreciation to A. Uezu,C. Banks-Leite, C. Cornellius and two anonymous reviewers forcommenting on earlier versions of this manuscript. Proof readingwas done by J.W. Reid. MCR was granted a CAPES scholarship,and additional financial support was provided by the BrazilianCouncil for Research and Technology (CNPq) (Project ‘Biodiversityconservation in fragmented landscapes on the Atlantic Plateau ofSão Paulo’, No. 590041/2006-1).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.biocon.2009.02.021.

References

Aleixo, A., 1999. Effects of a selective logging on a bird community in the BrazilianAtlantic Forest. The Condor 101, 537–548.

Antongiovanni, M., Metzger, J.P., 2005. Influence of matrix habitats on theoccurrence of insectivorous bird species in Amazonian forest fragments.Biological Conservation 122, 441–451.

Awade, M., Metzger, J.P., 2008. Using gap-crossing capacity to evaluate functionalconnectivity of two Atlantic rainforest birds and their response tofragmentation. Austral Ecology 33, 863–871.

Bacha, C.J.C., Barros, A.L.M.B., 2004. Reflorestamento no Brasil: evolução recente eperspectivas para o futuro. Scientia Forestalis 66, 191–203.

Baddeley, A., Turner, R., 2005. SPATSTAT: an R package for analysing spatial pointpattern. Journal of Statistical Software 12, 1–42.

Baptista, S.R., Rudel, T.K., 2006. A re-emerging Atlantic Forest? Urbanization,industrialization and the forest transition in Santa Catarina, southern Brazil.Environmental Conservation 33, 195–202.

Barlow, J., Gardner, T.A., Araujo, I.S., Avila-Pires, T.C., Bonaldo, A.B., Costa, J.E.,Esposito, M.C., Ferreira, L.V., Hawes, J., Hernandez, M.I.M., Hoogmoed, M.S.,Leite, R.N., Lo-Man-Hung, N.F., Malcolm, J.R., Martins, M.B., Mestre, L.A.M.,Miranda-Santos, R., Nunes-Gutjahr, A.L., Overal, W.L., Parry, L., Peters, S.L.,Ribeiro-Junior, M.A., Silva, M.N.F.da, Silva Motta, C., Peres, C.A., 2007a.Quantifying the biodiversity value of tropical primary, secondary, andplantation forests. Proceedings of the National Academy of Sciences 104,18555–18560.

Barlow, J., Mestre, L.A.M., Gardner, T.A., Peres, C.A., 2007b. The value of primary,secondary and plantation forests for Amazonian birds. Biological Conservation136, 212–231.

Baum, K.A., Haynes, K.J., Dillemuth, F.P., Cronin, J.T., 2004. The matrix enhances theeffectiveness of corridors and stepping stones. Ecology 85, 2672–2675.

Boscolo, D., Candia-Gallardo, C., Awade, M., Metzger, J.P., 2008. Importance of inter-habitat gaps and stepping-stones for a bird species in the Atlantic Forest, Brazil.Biotropica 40, 273–276.

Brannstrom, C., 2001. Conservation-with-development models in Brazil’s agro-pastoral landscapes. World Development 29, 1345–1359.

Brooks, T., Pimm, S.L., Oyugi, J.O., 1999. Time lag between deforestation and birdextinction in tropical forest fragments. Conservation Biology 13, 1140–1150.

Câmara, I.G., 2003. Brief history of conservation in the Atlantic Forest. In: Galindo-Leal, C., Câmara, I.G. (Eds.), The Atlantic Forest of South America: BiodiversityStatus, Threats, and Outlook. CABS and Island Press, Washington, pp. 31–42.

Cartes, J.L., Yanosky, A., 2003. Dynamics of biodiversity loss in the ParaguayanAtlantic Forest: an introduction. In: Galindo-Leal, C., Câmara, I.G. (Eds.), TheAtlantic Forest of South America: Biodiversity Status, Threats, and Outlook.CABS and Island Press, Washington, pp. 267–268.

Castellón, T.D., Sieving, K.E., 2005. An experimental test of matrix permeability andcorridor use by an endemic undestory bird. Conservation Biology 20, 135–145.

Cruz, C.B.M., Vicens, R.S., in press. Estado atual da cobertura vegetal e uso da terrano Bioma Mata Atlântica, in: Instituto Bioatlântica (Org.), Uma Introdução aoEstudo das Florestas do Sul e Sudeste do Brasil, Rio de Janeiro.

Cullen Junior, L., Fenimore, S., 2002. Projeto abraço verde: a practice-basedapproach to Brazilian Atlantic Forest conservation. Endangered SpeciesUpdate 10, 45–58.

Cullen Junior, L., Ferreira-Lima, J., Pavan, T., 2004. Agroforestry buffer zones andstepping stones: tools for the conservation of fragmented landscapes in theBrazilian Atlantic Forest. In: Schorth, G., Fonseca, G., Harvey, C., Claude, G.,Vasconcelos, H., Izac, A.N. (Eds.), Agroforestry and Biodiversity Conservation inTropical Landscapes. Island Press, Washington, pp. 415–430.

da Silva, J.M.C., Casteleti, C.H.M., 2003. Status of the biodiversity of the AtlanticForest of Brazil. In: Galindo-Leal, C., Câmara, I.G. (Eds.), The Atlantic Forest ofSouth America: Biodiversity Status, Threats, and Outlook. CABS and Island Press,Washington, pp. 43–59.

da Silva, J.M.C., de Souza, M.C., Castelletti, C.H.M., 2004. Areas of endemism forpasserine birds in the Atlantic Forest, South America. Global Ecology andBiogeography 13, 85–92.

Dean, W., 1996. With Broadax and Firebrand: The Destruction of the BrazilianAtlantic Forest. University of California Press, California.

Develey, P.F., Martensen, A.C., 2006. As aves da Reserva Florestal do morro Grande(Cotia, SP). Biota Neotropica 6 (2). <http://www.biotaneotropica.org.br/v6n2/pt/abstract?article+bn00706022006>.

Develey, P.F., Stouffer, P.C., 2001. Effects of roads on movements by understory birdsin mixed-species flocks in Central Amazonian Brazil. Conservation Biology 15,1416–1422.

Didham, R.K., Lawton, J.H., 1999. Edge structure determines the magnitude ofchanges in microclimate and vegetation structure in tropical forest fragments.Biotropica 31, 17–30.

Dunn, R.R., 2004. Recovery of faunal communities during tropical forestregeneration. Conservation Biology 18, 302–309.

Durigan, G., Siqueira, M.F., Franco, G.A.D.C., 2007. Threats to the Cerrado remnantsof the state of São Paulo, Brazil. Scientia Agricola 64, 363–366.

ESRI, 2007. ArcGis, the Complete Geographical Information System. Redlands,California.

Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity. Annual Review ofEcology, Evolution and Systematics 34, 487–515.

Fairbanks, D.H.K., Reyers, B., Van Jaarsveld, A.S., 2001. Species and environmentrepresentation: selecting reserves for the retention of avian diversity inKwaZulu-Natal, South Africa. Biological Conservation 98, 365–379.

Fall, A., Fortin, M.J., Manseau, M., OBrien, D., 2007. Spatial graphs: principles andapplications for habitat connectivity. Ecosystems 10, 448–461.

Ferraz, G., Nichols, J.D., Hines, J.E., Stouffer, P.C., Bierregaard Jr., R.O., Lovejoy, T.E.,2007. A large-scale deforestation experiment: effects of patch area and isolationon Amazon birds. Science 315, 238–241.

Fischer, J., Lindenmayer, D.B., Manning, A.D., 2006. Biodiversity, ecosystem function,and resilience: ten guiding principles for commodity production landscapes.Frontiers in Ecology and the Environment 4, 80–86.

Fonseca, C.R., Ganade, G., Baldissera, R., Becker, C.G., Boelter, C.R., Brescovit, A.D.,Campos, L.M., Fleck, T., Fonseca, V.S., Hartz, S.M., Joner, F., Käffer, M.I., Leal-Zanchet, A.M., Marcelli, M.P., Mesquita, A.S., Mondin, C.A., Paz, C.P., Petry, M.V.,

1152 M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153

Piovezan, F.N., Putzke, J., Stranz, A., Vergara, M., Vieira, E.M., 2009. Towards anecologically sustainable forestry in the Atlantic Forest. Biological Conservation142, 1144–1154.

Forman, R.T.T., Collinge, S., 1997. Nature conserved in changing landscapes with andwithout spatial planning. Landscape and Urban Planning 37, 129–135.

Fortin, M.J., Dale, M.R.T., 2005. Spatial Analysis: A Guide for Ecologists. CambridgeUniversity Press, Cambridge.

Galindo-Leal, C., Câmara, I.G., 2003. Atlantic Forest hotspot status: an overview. In:Galindo-Leal, C., Câmara, I.G. (Eds.), The Atlantic Forest of South America:Biodiversity Status, Threats and Outlook. CABS and Island Press, Washington,pp. 3–11.

Gardner, T.A., Barlow, J., Parry, L.W., Peres, C.A., 2007. Predicting the uncertainfuture of tropical forest species in a data vacuum. Biotropica 39, 25–30.

Gardner, T.A., Barlow, J., Araújo, I.S., Ávila-Pires, T.C., Bonaldo, A.B., Costa, J.E.,Esposito, M.C., Ferreira, L.V., Hawes, J., Hernandez, M.I.M., Hoogmoed, M.S.,Leite, R.N., Lo-Man-Hung, N.F., Malcolm, J.R., Martins, M.B., Mestre, L.A.M.,Miranda-Santos, R., Overal, W.L., Parry, L., Peters, S.L., Ribeiro-Junior, M.A., daSilva, M.N.F., Motta, C.S., Peres, C.A., 2008. The cost-effectiveness of biodiversitysurveys in tropical forests. Ecology Letters 11, 139–150.

Giraudo, A.R., 2003. Dynamics of biodiversity loss in the Argentinean AtlanticForest: an introduction. In: Galindo-Leal, C., Câmara, I.G. (Eds.), The AtlanticForest of South America: Biodiversity Status, Threats, and Outlook. CABS andIsland Press, Washington, pp. 139–140.

Goerck, J.M., 1997. Patterns of rarity in the birds of the Atlantic Forest of Brazil.Conservation Biology 11, 112–118.

Groves, C.R., Jensen, D.B., Valutis, L.L., Redford, K.H., Shaffer, M.L., Scott, J.M.,Baumgartner, J.V., Higgins, J.V., Beck, M.W., Anderson, M.G., 2002. Planning forbiodiversity conservation: putting conservation science into practice.Bioscience 52, 499–512.

Haddad, N.M., 1999. Corridor and distance effects on interpatch movements: alandscape experiment with butterflies. Ecological Applications 9, 612–622.

Haila, Y., 2002. A conceptual genealogy of fragmentation research: from islandbiogeography to landscape ecology. Ecological Applications 12, 321–334.

Hansbauer, M.M., Storch, I., Leu, S., Nieto-Holguin, J.P., Pimentel, R.G., Knauer, F.,Metzger, J.P., 2008. Movements of neotropical understory passerines affected byanthropogenic forest edges in the Brazilian Atlantic Rainforest. BiologicalConservation 141, 782–791.

Hanski, I., 1994. A pratical model of metapopulation dynamics. Journal of AnimalEcology 63, 151–162.

Harris, G.M., Pimm, S.L., 2004. Bird species’ tolerance of secondary forest habitatsand its effects on extinction. Conservation Biology 18, 1607–1616.

Harvey, C.A., Tucker, N.I.J., Estrada, A., 2004. Live fences, isolated trees, andwindbreaks: tools for conserving biodiversity in fragmented tropicallandscapes. In: Schorth, G., Fonseca, G., Harvey, C., Claude, G., Vasconcelos, H.,Izac, A.N. (Eds.), Agroforestry and Biodiversity Conservation in TropicalLandscapes. Island Press, Washington, pp. 261–289.

Huang, C., Kim, S., Altstatt, A., Townshend, J.R.G., Davis, P., Song, K., Tucker, C.J.,Rodas, O., Yanosky, A., Clay, R., Musinsky, J., 2007. Rapid loss of ParaguaysAtlantic forest and the status of protected areas – a landsat assessment. RemoteSensing of Environment 106, 460–466.

Instituto de Estudos Sócioambientais do Sul da Bahia (IESB), Instituto deGeociências da Universidade Federal do Rio de Janeiro (IGEO/UFRJ),Departamento de Geografia da Universidade Federal Fluminence (UFF), 2007.Levantamento da Cobertura Vegetal Nativa do Bioma Mata Atlântica. Relatóriofinal. PROBIO 03/2004, Brasília, 84p.

Jorge, L.A.B., Garcia, G.J., 1997. A study of habitat fragmentation in SoutheasternBrazil using remote sensing and geographic information systems (GIS). ForestEcology and Management 98, 35–47.

Kapos, V., 1989. Effects of isolation on the water status of forest patches in theBrasilian Amazon. Journal of Tropical Ecology 2, 173–185.

Landis, J.R., Kock, G.G., 1977. The measurement of observer agreement forcategorical data. Biometrics 33, 159–174.

Laurance, S.G.W., 2004. Responses of understory rain forest birds to road edges inCentral Amazonia. Ecological Applications 14, 1344–1357.

Laurance, W.F., 2007. Have we overstated the tropical biodiversity crisis? Trends inEcology and Evolution 22, 65–70.

Laurance, W.F., 2008. Theory meets reality: how habitat fragmentation research hastranscended island biogeographic theory. Biological Conservation 141, 1731–1744.

Laurance, S.G.W., Gómez, M.S., 2005. Clearing width and movements of understoryrainforest birds. Biotropica 37, 149–152.

Laurance, W.F., Nascimento, H.E.A.M., Laurance, S.G.W., Andrade, A., Ewers, R.M.,Harms, K.E., Luizão, R.C., Ribeiro, J.E.L.S., 2007. Habitat fragmentation, variableedge effects, and the landscape-divergence hypothesis. PLoS ONE 2,e1017.

Leica, 2006. Erdas Imagine 9.1, Leica Geosystems Geospatial Imaging, LLC.Lewinsohn, T.M., Prado, P.I., 2005. How many species are there in Brazil?

Conservation Biology 19, 619–624.Liebsch, D., Marques, M.C.M., Goldenberg, R., 2008. How long does the Atlantic Rain

Forest take to recover after a disturbance? Changes in species composition andecological features during secondary succession. Biological Conservation 141,1717–1725.

Lindborg, R., Eriksson, O., 2004. Historical landscape connectivity affects presentplant species diversity. Ecology 85, 1840–1845.

Lindenmayer, D.B., Franklin, J.F., Fischer, J., 2006. General management principlesand a checklist of strategies to guide forest biodiversity conservation. BiologicalConservation 131, 433–445.

Lindenmayer, D., Hobbs, R.J., Montague-Drake, R., Alexandra, J., Bennett, A.,Burgman, M., Cale, P., Calhoun, A., Cramer, V., Cullen, P., Driscoll, D., Fahrig, L.,Fischer, J., Franklin, J., Haila, Y., Hunter, M., Gibbons, P., Lake, S., Luck, G.,MacGregor, C., McIntyre, S., Mac Nally, R., Manning, A., Miller, J., Mooney, H.,Noss, R., Possingham, H., Saunders, D., Schmiegelow, F., Scott, M., Simberloff, D.,Sisk, T., Tabor, G., Walker, B., Wiens, J., Woinarski, J., Zavaleta, E., 2008. Achecklist for ecological management of landscape for conservation. EcologyLetters 11, 78–91.

Lopes, A.V., Girão, L.C., Santos, B.A., Peres, C.A., Tabarelli, M., 2009. Long-termerosion of tree reproductive trait diversity in edge-dominated Atlantic forestfragments. Biological Conservation 142, 1154–1165.

Margules, C.R., Pressey, R.L., 2000. Systematic conservation planning. Nature 405,243–253.

Martensen, A.C., Pimentel, R.G., Metzger, J.P., 2008. Relative effects of fragment sizeand connectivity on bird community in the Atlantic Rain Forest: implicationsfor conservation. Biological Conservation 141, 2184–2192.

Metzger, J.P., 2000. Tree functional group richness and landscape structure in aBrazilian tropical fragmented landscape. Ecological Applications 10, 1147–1161.

Metzger, J.P., 2006. How to deal with non-obvious rules for biodiversityconservation in fragmented areas. The Brazilian Journal of NatureConservation 4, 125–137.

Metzger, J.P., Ribeiro, M.C., Ciocheti, G., Tambosi, L.R., 2008. Uso de índices depaisagem para a definição de ações de conservação e restauração dabiodiversidade do Estado de São Paulo. In: Rodrigues, R.R., Joly, C.A., Brito,M.C.W., Paese, A., Metzger, J.P., Casatti, L., Nalon, M.A., Menezes, N., Ivanauskas,N.M., Bolzani, V., Bononi, V.L.R. (Eds.), Diretrizes para Conservação eRestauração da Biodiversidade no Estado de São Paulo. Secretaria do MeioAmbiente and Fapesp, São Paulo, pp. 120–127.

Metzger, J.P., Martensen, A.C., Dixo, M., Bernacci, L.C., Ribeiro, M.C., Teixeira, A.M.G,Pardini, R., 2009. Time-lag in biological responses to landscape changes in ahighly dynamic Atlantic forest region. Biological Conservation 142, 1166–1177.

Minor, E.S., Urban, D., 2007. A graph-theory framework for evaluating landscapeconnectivity and conservation planning. Conservation Biology 17, 1771–1782.

Mittermeier, R.A., Myers, N., Robles Gil, P., Mittermeier, C.C., 1999. Hotspots.Agrupación Sierra Madre, CEMEX, Mexico City.

MMA (Ministério do Meio Ambiente), 2007. <www.mma.gov.br> .Murcia, C., 1995. Edge effects in fragmented forests: implications for conservation.

Tree Review 10, 58–62.Nassar, A.M., Rudorff, B.F.T., Antoniazzi, L.B., Aguiar, D.A., Bacchi, M.R.P., Adami, M.,

2008. Prospects of the sugarcane expansion in Brazil: impacts on direct andindirect land use changes. In: Zuurbier, P., Van de Vooren, J. (Eds.), SugarcaneEthanol: Contribution to Climate Change Mitigation and the Environment.Wageningen Academic Publishers, The Netherlands, 2008, cap. 3, pp. 63-93.

Neteler, M., Mitasova, H., 2008. Open source GIS: a GRASS GIS approach, third ed..The International Series in Engineering and Computer Science, vol. 773Springer, New York. 406p.

O’Brien, D., Manseau, M., Fall, A., Fortin, M.J., 2006. Testing the importance of spatialconfiguration of winter habitat for woodland caribou: an application of graphtheory. Biological Conservation 130, 70–83.

Oliveira-Filho, A.T., Fontes, M.A.L., 2000. Patterns of floristic differentiation amongAtlantic Forests in Southeastern Brazil and the influence of climate. Biotropica34, 793–810.

Pardini, R., Faria, D., Accacio, G.M., Laps, R.R., Mariano, E., Paciencia, M.L.B., Dixo, M.,Baumgarten, J., 2009. The challenge of maintaining Atlantic forest biodiversity:a multi-taxa conservation assessment of an agro-forestry mosaic in southernBahia. Biological Conservation 142, 1178–1190.

Parker III, T.A., Stotz, D.F., Fitzpatrick, J.W., 1996. Ecological and distributionaldatabases. In: Stotz, D.F., Fitzpatrick, J.W., Parker, T.A., III, Moskovits, D.K. (Eds.),Neotropical Birds: Ecology and Conservation. University of Chicago Press,Illinois, pp. 131–146.

R Development Core Team, 2008. R: a Language and Environment for StatisticalComputing. R Foundation for Statistical Computing, Vienna, Austria (<http://www.R-project.org>).

Ranta, P., Blom, T., Niemelä, J., Joensuu, E., Siitonen, M., 1998. The fragmentedAtlantic rain forest of Brazil: size, shape and distribution of forest fragments.Biodiversity and Conservation 7, 385–403.

Ricketts, T.H., 2001. The matrix matters: effective isolation in fragmentedlandscapes. American Naturalist 158, 87–99.

Rodrigues, R.R., Lima, R.A.F., Gandolfi, S., Nave, A.G., 2009. On the restoration of highdiversity forests: 30 years of experience in the Brazilian Atlantic Forest.Biological Conservation 142, 1242–1251.

Rudorff, B.F.T., Sugawara, L.M., 2007. Mapeamento da cana-de-açúcar na RegiãoCentro-Sul via imagens de satélites. Informe Agropecuário 28, 79–86.

Secretariat of the Convention on Biological Diversity, 2002. Global Strategy for PlantConservation. Secretariat of the Convention on Biological Diversity. CBD, UMEP,Botanical Gardens Comservation International, Montreal.

Sekercioglu, C.H., Loarie, S.R., Brenes, F.O., Ehrlich, P.R., Daily, G.C., 2006. Persistenceof forest birds in the Costa Rican agricultural countryside. Conservation Biology21, 482–494.

M.C. Ribeiro et al. / Biological Conservation 142 (2009) 1141–1153 1153

Silva, W.G.S., Metzger, J.P., Simões, S., Simonetti, C., 2007. Relief influence on thespatial distribution of the Atlantic Forest cover at the Ibiúna Plateau, SP.Brazilian Journal of Biology 67, 403–411.

SOS Mata Atlântica, Instituto Nacional de Pesquisas Espaciais, 2000. Atlas dosRemanescentes Florestais e Ecossistemas Associados no Domínio da MataAtlântica. São Paulo.

SOS Mata Atlântica, Instituto Nacional de Pesquisas Espaciais, 1993. Atlas daevolução dos remanescentes florestais da Mata Atlântica e ecossistemasassociados no período de 1985–1990. São Paulo.

SOS Mata Atlântica, Instituto Nacional de Pesquisas Espaciais, 2008. Atlas dosremanescentes florestais da Mata Atlântica, período de 2000 a 2005. <http://www.sosmatatlantica.org.br>.

Stotz, D.F., Fitzpatrick, J.W., Parker, T.A., Moskovitz, D.K., 1996. Neotropical Birds:Ecology and Conservation. Univ. Chicago Press, Chicago.

Teixeira, A.M.G., Soares-Filho, B.S., Freitas, S.R., Metzger, J.P., 2009. Modelinglandscape dynamics in an Atlantic Rainforest region: implications forconservation. Forest Ecology and Management 257, 1219–1230.

Uehara-Prado, M., Fernandes, J.O., Bello, A.M., Machado, G., Santos, A.J., Vaz-de-Mello, F.Z., Freitas, A.V.L., 2009. Selecting terrestrial arthropods as indicators ofsmall-scale disturbance: a first approach in the Brazilian Atlantic forest.Biological Conservation 142, 1220–1228.

Uezu, A., Metzger, J.P., Vielliard, J.M.E., 2005. Effects of structural and functionalconnectivity and patch size on the abundance of seven Atlantic Forest birdspecies. Biological Conservation 123, 507–519.

Uezu, A., Beyer, D.D., Metzger, J.P., 2008. Can agroforest woodlots work as steppingstones for birds in the Atlantic Forest region? Biodiversity and Conservation 17,1907–1922.

Umetsu, F., Pardini, R., 2007. Small mammals in a mosaic of forest remnants andanthropogenic habitats: evaluating matrix quality in an Atlantic forestlandscape. Landscape Ecology 22, 517–530.

Umetsu, F., Metzger, J.P., Pardini, R., 2008. The importance of estimating matrixquality for modeling species distribution in complex tropical landscape: a testwith Atlantic forest small mammals. Ecography 31, 359–370.

Urban, D.L., 2005. Modeling ecological processes across scales. Ecology 86, 1996–2006.

Urban, D., Keitt, T., 2001. Landscape connectivity: a graph-theoretic perspective.Ecology 82, 1205–1218.

Viana, V.M., Tabanez, A.A., 1996. Biology and conservation of forest fragments in theBrazilian Atlantic moist forest. In: Schelhas, J., Greenberg, R. (Eds.),Forest Patches in Tropical Landscapes. Island press, Washington, pp. 151–167.

Viana, V.M., Tabanez, A.A.J., Batista, J.L., 1997. Dynamic and restoration of forestfragments in the Brazilian Atlantic moist forest. In: Laurance, W., Bierregaard,R., Jr. (Eds.), Tropical Forest Remnants: Ecology, Management, and Conservationof Fragmented Communities. The University of Chicago Press, Chicago andLondon, pp. 351–365.

Vieira, M.V., Olifiers, N., Delciellos, A.C., Antunes, V.Z., Bernardo, L.R., Grelle, C.E.V.,Cerqueira, R., 2009. Land use vs. fragment size and isolation as determinants ofsmall mammal composition and richness in Atlantic Forest remnants. BiologicalConservation 142, 1191–1200.

Vos, C.C., Verboom, J., Opdam, P.F.M., Ter Braak, C.J.F., 2001. Toward ecologicallyscaled landscape indices. American Naturalist 183, 24–41.

Wilcox, B.A., Murphy, D.D., 1985. Conservation strategy: effects of fragmentation onextinction. The American Naturalist 125, 879–887.

Williams, P.H., Margules, C.R., Wilbert, D.W., 2002. Data requirements and datasources for biodiversity priority area selection. Journal of Bioscience 27, 327–338.

Wright, S.J., 2005. Tropical forests in a changing environment. Trends in Ecology andEvolution 20, 553–560.